GOMS Model: Predicting User Performance as a Script

WHY IT EXISTS Before the 1980s, there was no formal way to predict if a user interface would be efficient before building and testing it. GOMS was created to bring an engineering discipline to Human-Computer Interaction (HCI), allowing designers to quantify and predict user performance for routine tasks, much like an engineer predicts the performance of a physical machine.

THE MENTAL MODEL Think of an expert user as a computer executing a program. The user has a Goal (the desired outcome), and to achieve it, they run a Method (a pre-learned procedure). This method is composed of a sequence of Operators (the smallest physical or mental actions, like moving a mouse or pressing a key). If multiple methods exist, a Selection rule (an if-then statement) determines which one to use based on context.

HOW IT WORKS A GOMS analysis involves four components. First, define the user's high-level Goals. Second, list the available Methods to achieve those goals. Third, break down each method into its constituent Operators—the primitive actions like 'move hand to mouse' or 'press key'. Each operator has a standardized, empirically derived time cost. Fourth, define Selection rules for when a user would choose one method over another, usually based on efficiency. By summing the times of the operators in a chosen method, you can predict the total task time for an expert user.

WHEN TO USE IT GOMS is best for analyzing and optimizing existing, highly repetitive tasks performed by expert users where efficiency is paramount. For example, comparing two keyboard shortcut layouts for a data entry clerk or predicting the time saved by a new feature in a call center application. It provides quantitative data to justify design decisions before implementation.

WHEN NOT TO USE IT Do not use GOMS to model novice users, as it doesn't account for learning or exploration. It's also inappropriate for creative or problem-solving tasks that don't have a predefined, optimal procedure. Because it assumes error-free performance, it cannot predict where users will make mistakes or how they will recover from them.

ONE CANONICAL EXAMPLE Consider the goal of deleting a word in a text editor. Method 1: Move hand to mouse, move cursor to start of word, click and drag to end of word, move hand to keyboard, press Delete. Method 2: Use keyboard shortcuts to move the cursor to the word, then use a key combination to delete it. A GOMS analysis would assign time values to each operator (move hand, click, press key) and sum them up for each method to predict which is faster for an expert typist.